(gdb) frame
#0  0x00007f8262bc6a8e in sysdb_msg2attrs (mem_ctx=mem_ctx@entry=0x560a69714c30, count=277, msgs=0x560a69744bc0, attrs=attrs@entry=0x7ffd9c9ada80)
    at src/db/sysdb.c:1574
1574	        a[i]->num = msgs[i]->num_elements;
(gdb)  
(gdb) p count
$11 = 277
(gdb) p i
$12 = 152
(gdb) p msgs[i]
$13 = (struct ldb_message *) 0x0


(gdb) frame 1
#1  0x0000560a67f14670 in be_refresh_get_values_ex (_values=0x560a69714f78, search_cache=SYSDB_CACHE_TYPE_TIMESTAMP, value_attr=0x7f82510d430b "name", 
    key_attr=<optimized out>, base_dn=0x560a69601b40, period=4050, domain=0x560a695c44e0, mem_ctx=0x560a69714f40) at src/providers/be_refresh.c:75
75	    ret = sysdb_msg2attrs(tmp_ctx, res->count, res->msgs, &records);
(gdb)  
(gdb) p res->msgs[151]
$26 = (struct ldb_message *) 0x560a69811220
(gdb) p res->msgs[152]
$27 = (struct ldb_message *) 0x0
(gdb) p res->msgs[153]
$28 = (struct ldb_message *) 0x0
(gdb) p res->msgs[154]
$29 = (struct ldb_message *) 0x560a69814190
(gdb) p res->msgs[155]
$30 = (struct ldb_message *) 0x560a69814c40


  --  `sysdb_search_with_ts_attr()` returned `ldb_result` with some of msgs[] being NULL ptr.


sysdb_search_with_ts_attr(SYSDB_CACHE_TYPE_TIMESTAMP) = sysdb_search_ts_entry() + merge_res_sysdb_attrs(),
   merge_res_sysdb_attrs() = for (c:count) { merge_msg_sysdb_attrs(msgs[c]) }
      merge_msg_sysdb_attrs() = merge_all_ts_attrs(sysdb_cache_search_entry())

But if `sysdb_cache_search_entry()` fails then `merge_res_sysdb_attrs()` doesn't bother much and `ts_cache_res->msgs[c]` is left zero initialized:
https://github.com/SSSD/sssd/blob/d93b4fe14b0f72bd8311497d18204f153c104007/src/db/sysdb_search.c#L252
(btw, I think comment is wrong here)

But if my guess is correct we should see "Cannot merge sysdb cache values for %s\n" SSSDBG_MINOR_FAILURE messages in the log.

I didn't find sssd logs in the sos-report.


Akshay, can we get sssd_$domain.log please?

Btw, if they will remove cache db most probably issue will be hidden. Would be good to get log before cache removed. Or are they able to re-produce the issue reliable?

Pushed PR: https://github.com/SSSD/sssd/pull/5414

* `master`
    * ff24d1538af88f83d0a3cc2817952cf70e7ca580 - SYSDB: merge_res_sysdb_attrs() fixed to avoid NULL ptr in msgs[]
* `sssd-1-16`
    * 9ace3a7899e6b3753ef428088303e0a646db4096 - SYSDB: merge_res_sysdb_attrs() fixed to avoid NULL ptr in msgs[]

We could also reproduce this with the AD backend on 7.9.
The linked patch seems to fix the issue (we have sssd running for 20 minutes now, whereas it previously crashed after 1-2 minutes).

Although we had to build packages with the fixes locally anyway, updated official packages would be much appreciated. This issue has the potential to break logins for all AD-integrated machines.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (sssd bug fix and enhancement update), and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2020:5459